Predicting real-time deformation of structure in fire using machine learning with CFD and FEM

Abstract

Real-time prediction of structural safety conditions is critical to firefighting teams during building fire rescue operations. This paper presents a numerical data-based machine learning (ML) framework for structural fire response prediction by integrating Computational Fluid Dynamics (CFD) and Finite Element Methods (FEM). Multiple ML models are developed to predict the real-time vertical displacement of the structure based on corresponding temperature field data with a numerical case of an 8 m × 8 m × 0.6 m virtual steel roof structure under 1200 virtual fire scenarios. The models are comparatively evaluated in terms of predictive accuracy, computational efficiency, and model robustness. The results show the effectiveness of Random Forest and Gradient Boosting models in real-time displacement prediction, indicating the feasibility of the framework in providing timely and reliable predictions of structural safety conditions for firefighters. The CFD/FEM-based framework could be embedded in a structural fire safety warning system to facilitate smart firefighting in the future.